Hello list,
I am having a look at some PET data, where we had four conditions
(Task A and J combined with stimulus O and W -> conditions AO, AW, JO,
JW). 14 subjects were scanned with no baseline. Each condition was
repeated three times (=12 scans/subject). I analyzed single subjects
first and performed a group analysis as a 2nd level design.
On the first level, tasks and stimuli were orthogonal, i.e. the
timecourse of task A was -(timecourse of task J). Thus, i created two
EVs, EV1 for task A and EV2 for stimulus O. Main effects were defined
as:
EV1 EV2
1 0 task A
-1 0 task J
0 1 stimulus O
0 -1 stimulus W
Results are very reasonable for the main effects. However, I am more
interested in the interaction of task A and stimulus O, i.e. the
neuronal correlates of task A particularly when triggered by stimulus
O. I therefore added a third interaction EV encompassing contrasts 1
and 3 [1 0; 0 1], setting "make zero: min" for both. The resulting
regressor is "on" only for the AO-condition, as intended (see attached
design matrix image). Results are in a anatomically plausible and
expected region. However, and this is my problem, the COPEs given by
Featquery are just the other way round as I imagined them to be! In
this region/z-thresholded mask (and for the max peak) they are
something like -300 +300 -420 +420 (for A, J, O and W, respectively) -
that is, just the negative of the expected pattern. I can't get my
head around this - can anyone help me out of this?
Many thanks in advance and best regards,
Cornelius
--
Dr. med. Cornelius J. Werner
Department of Neurology
RWTH Aachen University
Pauwelsstr. 30
52074 Aachen
Germany
|